{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.02440", "kind": "paper", "url": "https://arxiv.org/abs/2607.02440"}, {"id": "gh:linzwcs/evopolicygym", "kind": "repo", "url": "https://github.com/Linzwcs/EvoPolicyGym"}], "authors": ["Zhilin Wang", "Han Song", "Runzhe Zhan", "Jusen Du", "Jiacheng Chen", "Tianle Li", "Qingyu Yin", "Yulun Wu", "Zhennan Shen", "Tong Zhu", "Yanshu Li", "Guanjie Chen", "Derek F. Wong", "Yafu Li", "Yu Cheng", "Yang Yang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.02440", "source_url": "https://arxiv.org/abs/2607.02440"}, "categories": ["Agents & Tool Use", "General AI", "cs.AI"], "description": {"en": "EvoPolicyGym evaluates autonomous policy evolution in interactive RL environments. A harness-model agent iteratively edits an executable policy under a fixed interaction budget. The benchmark records trajectories of programs, submissions, feedback, and selection, and scores agents on held-out episodes."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.02440", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.02440", "title": "EvoPolicyGym"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of Science and Technology of China", "Chinese University of Hong Kong", "University of Macau", "Tsinghua University", "Zhejiang University", "Shanghai Jiao Tong University", "Soochow University", "Brown University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.02440", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.02440", "source_url": "https://openalex.org/W7167239711"}, "key": "claire-radar:2607.02440", "modality": null, "name": "EvoPolicyGym", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_evopolicygym_6d1d11a9", "origin_source": "arxiv", "recognition_confidence": "0.85", "record_sha256": "070812025710671ff23895c4062625cacf6cae47a5df5397cfabdd5850f27f7f", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.02440"}, "publisher": null, "released": "2026-07-02", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.02440", "source_url": "https://arxiv.org/abs/2607.02440"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-02440", "source": "claire_radar", "source_benchmark_id": "2607.02440", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Mixed", "applicationDomains": ["General AI"], "area": "Agents & Tool Use", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/Linzwcs/EvoPolicyGym", "githubScope": "benchmark_repo", "githubStars": 175, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 48, "hfPaperUrl": "https://huggingface.co/papers/2607.02440", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-10-04", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "Medium", "construction": "Aggregate Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "a657c35593f2b602c149c383fb8440215e4e327a769b92ed67e2939bb4795725", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "EvoPolicyGym evaluates autonomous policy evolution in interactive RL environments. A harness-model agent iteratively edits an executable policy under a fixed interaction budget. The benchmark records trajectories of programs, submissions, feedback, and selection, and scores agents on held-out episodes.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "evaluation protocol evidence"], "snippet": "We instantiate this setting in EvoPolicyGym, a benchmark built from compact interactive RL environments that evaluates how agents iteratively improve explored policies."}, "familyId": "bmf_887be45686cc", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/Linzwcs/EvoPolicyGym", "url": "https://github.com/Linzwcs/EvoPolicyGym"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/Linzwcs/EvoPolicyGym", "url": "https://huggingface.co/papers/2607.02440"}, {"basis": "huggingface-paper-association", "linkedUrl": "https://github.com/Linzwcs/EvoPolicyGym", "url": "https://huggingface.co/api/papers/2607.02440"}], "scope": "benchmark_repo", "url": "https://github.com/Linzwcs/EvoPolicyGym", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_evopolicygym_6d1d11a9", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/Linzwcs/EvoPolicyGym", "data": null, "hfPaper": "https://huggingface.co/papers/2607.02440", "paper": "https://arxiv.org/abs/2607.02440", "pdf": "https://arxiv.org/pdf/2607.02440", "project": null, "report": "https://arxiv.org/abs/2607.02440"}, "metrics": [], "motivation": "Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress.", "name": "EvoPolicyGym", "oneLine": "EvoPolicyGym evaluates autonomous policy evolution in interactive RL environments. A harness-model agent iteratively edits an executable policy under a fixed interaction budget. The benchmark records trajectories of programs, submissions, feedback, and selection, and scores agents on held-out episodes.", "paperTitle": "EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.02440", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "EvoPolicyGym Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/Linzwcs/EvoPolicyGym"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 0.85, "relation": "introduces", "releasedAt": "2026-07-02", "source": {"authors": ["Zhilin Wang", "Han Song", "Runzhe Zhan", "Jusen Du", "Jiacheng Chen", "Tianle Li", "Qingyu Yin", "Yulun Wu", "Zhennan Shen", "Tong Zhu", "Yanshu Li", "Guanjie Chen", "Derek F. Wong", "Yafu Li", "Yu Cheng", "Yang Yang"], "categories": ["cs.AI", "cs.CL"], "id": "2607.02440", "title": "EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments", "type": "arxiv", "url": "https://arxiv.org/abs/2607.02440"}, "sourceUpdatedAt": "2026-07-02T17:10:13Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Existing evaluations often collapse iterative improvement into a final score or confound it with software-engineering progress. EvoPolicyGym isolates the capability to improve policies from bounded feedback, providing trajectory-level diagnostics that distinguish how agents allocate budget and refine policies. This supports comparison of agents on a controlled, reusable protocol."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
